Antimicrobial resistance (AMR) has become a tremendous global public health issue, which affects humans, animals and the environment. One of the major threats today is the worldwide dissemination of multi-drug resistant bacteria, particularly the extended-spectrum beta-lactamase (ESBL)-producing Enterobacterales (ESBL-E) (Lynch et al., 2013). Indeed, since the 2000s the incidence of ESBL-E has been steadily increasing in the community especially in middle- and low-income countries (Woerther et al., 2013). A meta-analysis conducted in 2016 estimated that 14% of healthy individuals worldwide were carriers of ESBL-E (Karanika et al., 2016). However, large regional disparities can be observed, with an estimated prevalence of 5% in Europe, 20% around the Mediterranean basin and in Africa, 50% in the Pacific and 70% in South-East Asia (Woerther et al., 2011; Karanika et al., 2016). An important cornerstone in the control of AMR is the implementation and monitoring of AMR surveillance indicators [World Health Organization [WHO], 2015]. In this context, the World Health Organization (WHO) proposed a surveillance protocol called Tricycle aiming to provide a simplified, integrated and trans-sectoral surveillance system. The proposed surveillance focuses on a single key indicator, the prevalence of ESBL-Escherichia coli, in the three major settings that are the human (community carriage and hospital infection), the food-chain (poultry) and the environment (water). E. coli was chosen as the target organism because it is a commensal of the gut of all humans and a wide variety of warm-blooded animals, and is also widespread in the environment (Savageau, 1983), but primarily because infections with ESBL-E. coli in humans, result in severe morbidity and mortality (Schwaber and Carmeli, 2007). For the human component of the Tricycle protocol, the population targeted as representative of the community was healthy pregnant women approaching or at delivery.

Thanks to a structured laboratory network (RESAMAD), the Tricycle protocol was implemented in Madagascar in 2018 by Fondation Mérieux to support the implementation of a functional National Action Plan on AMR. The need for such a surveillance system in the country was urgent as data on AMR were sparce and the few studies conducted previously reported increasing ESBL-E rates in the community, going from 10.1% in 2010 (Herindrainy et al., 2011) to 18.5% in 2014 (Chereau et al., 2015).

Here we report the first results from the human community component of the Tricycle protocol in Madagascar, describing ESBL-E. coli carriage and its associated risk factors among healthy pregnant women. We also investigated the genomic characteristics of isolated strains such as presence of resistance and virulence genes, plasmids and their genomic diversity.

Thanks to a structured network of laboratories (RESAMAD), we were able to conduct a large-scale study and then report the first results of the human community component of the Tricycle protocol in Madagascar. Our main results highlight a very high ESBL-E. coli carriage rate in healthy pregnant women both in the capital and in province with significant variations related to seasonality. We noted a high genomic diversity of strains carrying the bla_CTX–M–15 ESBL gene and belonging mostly to phylogenetic group A.

We chose to conduct the present study in Madagascar because of its specific characteristics. Besides the fact that little is known about ESBL-E in the country, it is a large island in the Indian ocean with a unique ecosystem and an important crossroads between Africa and Asia with significant immigration and commercial exchanges between both continents.

We observed an ESBL-E. coli prevalence in pregnant women of 34%, ranging from 12% to 65% depending on the sampling site. Unlike previous studies that focused mainly on the capital Antananarivo and nearby areas, we included three more cities in the provinces to depict a more comprehensive picture of ESBL-E dissemination in the country. Two previous studies conducted in Madagascar, evaluated community carriage of ESBL-E in Antananarivo to 10.1% in 2009 (Herindrainy et al., 2011) and to 18.5% in 2015 in Antananarivo and adjacent areas (Chereau et al., 2015). Our study underlines the worsening problem of ESBL dissemination in Madagascar as ESBL-E. coli alone was isolated from 34% of the participants which is much higher than previous reports. These results place Madagascar among the countries with the highest rates of ESBL prevalence confirming the global trends of ESBL epidemiology (Bezabih et al., 2021). This high carriage rate of ESBL-E. coli increases the risk of infection caused by ESBL-E. coli and, because of probabilistic treatment failures, triggers the use of broad-spectrum beta-lactams such as carbapenems which in its turn drives the spread of carbapenem resistant strains.

The main risk factor for ESBL-E. coli carriage that we identified was the wet season, with increased prevalence when rainfall and temperatures are higher. We observed that risk of carriage increased 1.4 times by 100 mm of rainfall and 1.1 times by 1°C of temperature increase. These results are robust with respect to center heterogeneity as shown by the sensitivity analysis with logistic mixed models. Even if this correlation has been described previously in Madagascar (Chereau et al., 2015) as well as in other parts of the world (MacFadden et al., 2018; McGough et al., 2020; Wielders et al., 2020), in light of the large number of scientific reports on ESBL-E carriage, the climate-related risk factor has rarely been reported. This result could be explained by different factors. First the rainy period is well known as being favorable for fecal-oral route of infectious pathogens transmission and enteric bacteria outbreaks (Chowdhury et al., 2018; Chao et al., 2019). In addition, higher prevalence of ESBL-E. coli in the rainy and warmer season could further increase horizontal genetic exchanges. It is well established in in vitro models that higher growth rates and elevated temperature and humidity are closely related (Ratkowsky et al., 1982) and could promote horizontal gene transfer (Hashimoto et al., 2019). Therefore, one could expect an even worse situation in the future with climate change and global warming trends (Ukuhor, 2021). We found no clear link with socio-economic factors which suggests that ESBL-E. coli disseminates evenly in the community in Madagascar mediated by environmental and person to person exchanges. Obviously other factors such as environmental and food contamination may play an important role in the high prevalence of ESBL-E carriage in the human community. It is also important to note that ESBL-E colonizing the gut microbiota of pregnant women can be transmitted to the child during delivery and thus increase the risk of severe ESBL-E infections in newborns (Danino et al., 2018; Herindrainy et al., 2018). Indeed, the incidence rate of severe neonatal infections with multidrug-resistant bacteria (mainly ESBL-E) was estimated at 5 cases per 1,000 live births in three developing countries, including Madagascar. Unfortunately, the origin of the bacteria responsible for the infection (mother, family members, and hospital acquisition) was not investigated (Huynh et al., 2021).

The most frequent ESBL gene identified was by far bla_CTX–M–15 which is consistent with global ESBL epidemiology (Bevan et al., 2017). This is probably due to successful E. coli strains carrying bla_CTX–M–15 on IncF plasmids, known for their persistence even in the absence of selective pressure (Mahérault et al., 2019). A high proportion of blaCTX-M genes (32%) was predicted to be chromosomic which is in agreement with previous studies (Hamamoto and Hirai, 2019). Chromosomic location provides stabilization of the blaCTX-M gene without additional cost, unlike a plasmid, and could help maintain resistance in the environment, even without selection pressure (Yoon et al., 2020; Shawa et al., 2021). Given that the protocol of the study was not designed to specifically target carbapenemase producing E. coli, isolation of an NDM-5 producing strain suggests a spread of carbapenemase producing Enterobacterales in the community of Mahajanga. To our knowledge, this is the first description of NDM-5 metalo beta-lactamase in Madagascar. Carbapenems are available in the country but their use is limited because of their high price (Herindrainy et al., 2011). Therefore, isolation of a strain carrying bla_NDM–5 in the absence of selective pressure is extremely worrying.

The molecular analysis of the isolates revealed a great variety of strains in all the sampling locations which was also observed in other developing countries (Stoesser et al., 2015; Ramadan et al., 2020; Foster-Nyarko et al., 2021). The majority were from commensal phylogenetic groups A with STc10 being the most frequent clonal complex detected. The second most frequent was STc155 (also named CC87 using the Pasteur Institute MLST scheme) with isolates from this STc belonging mainly to ST58 and ST155. STc155 strains are considered of animal origin with an evolution from wild type strains, becoming progressively resistant and, in many cases, later acquiring virulence genes (Skurnik et al., 2016). Indeed, in recent years multi-drug resistant (MDR) STc155 strains have been isolated with increasing frequency from extra-intestinal infections in humans (Royer et al., 2021). In our study STc155 isolates carried only a limited number of virulence genes but had already acquired an MDR profile. This finding underlines the importance of animal to human transmission in the dissemination of ESBL-E. coli in the community. Our results are consistent with previous reports describing higher prevalence of phylogenetic group A isolates in developing countries in contrast to industrialized countries where phylogenetic group B2 isolates are the most frequent (Duriez et al., 2001; Tenaillon et al., 2010). The phylogenetic tree based on a core genome SNP analysis showed a classical distribution of isolates by phylogenetic group except for a subgroup of phylogenetic group A isolates clustering between isolates from phylogenetic groups D and C. This subgroup contains STs found in very limited numbers in the EnteroBase database, which however, includes over 184,000 E. coli genomes. This indicates the potential emergence of a distinct sub-group beginning to diverge genetically from phylogenetic group A. As expected, we found few virulence genes belonging mainly to the adhesins group and a small number of toxins which is consistent with the commensal nature of most of the isolates. Even though most isolates belonged to commensal phylogenetic groups and therefore are less likely to cause invasive disease, infections with commensal strains are not uncommon (Leverstein-van Hall et al., 2011; Dadi et al., 2020). Also, one can imagine that these ESBL-E. coli strains can reside for a long time in the gut microbiota as they are well adapted to their host and thus represent an significant reservoir of resistance genes and a potential source that drives human to human transmission (Tande et al., 2010).

To analyze more precisely ESBL-E. coli dissemination we focused on isolates differing by less than 10 SNPs, a cut-off used in the literature to define closely related isolates (Schürch et al., 2018). Core-genome based SNP analysis revealed a limited number of closely related isolates circulating in each sampling location, which indicates probable inter-human transmission or a common contamination origin. Isolates differing from each other by less than 10 SNPs, especially from ST450 and ST361 have also been observed in different locations. In addition, most inter-city interactions were linked to the capital Antananarivo and to a lesser extent to provincial cities only. Assuming a mutation rate of 1.1 SNP per genome and per year as described elsewhere (Reeves et al., 2011) and equal mutation rates for all isolates, this finding suggests recent inter-city dissemination events probably due to traveling with journeys to the capital playing a central role in dissemination. Another hypothesis is that some clones, may have lower natural mutation rates than others.

Plasmid content analysis revealed an unusually prevalent IncY plasmid compared to other studies (Lyimo et al., 2016; Salinas et al., 2019), suggesting a dissemination in the country of an IncY type plasmid harboring a resistance genes island containing bla_CTX–M–15 and six other resistant genes of different antibiotic classes. To our knowledge, this is the first report of such a high prevalence of an IncY plasmid in ESBL-E. coli isolates. The extremely high sequence identity shared with an IncY plasmid in an E. coli strain isolated in Finland from Branta leucopsis (Kurittu et al., 2021) indicates that migratory birds could represent a powerful means of resistance spread even if a direct link to E. coli isolates from Madagascar is unlikely, due to the Northern Hemisphere habitat of this bird species (Owen and Black, 1989).

The present study has several limitations. First, due to logistic constraints, the sampling in the different hospitals was not performed during the same period. This could introduce a bias as risk factors may differ in different seasons. However, we performed a sensitivity analysis with logistic mixed models that addresses this bias. Second, several samples were eliminated because of transportation issues, but we were still able to perform phenotypic and genotypic analysis on 92% of expected isolates. Third, the Tricycle protocol requires three isolates per pregnant woman but certain participant hospitals did not collect all the three. Moreover, in some cases less than three presumptive ESBL-E. coli colonies were present after sample culture. In all, we were able to obtain three isolates in 42% of the participants. Moreover, even when successfully obtained, these three isolates often represent the dominant colonizing strains colonizing and one cannot exclude the existence of extensive in-host genetic diversity among non-dominant bacterial populations. Finally, we were able to sequence by the Nanopore technology only 12 out of 38 isolates carrying IncY plasmid due to budget limitations but we tried to compensate for this by aligning the Illumina reads of the remaining isolates to the IncY sequences reconstructed from the Nanopore sequencing.

We report here that in 2018–2019 more than one third of pregnant women in four study cities in Madagascar were ESBL-E. coli carriers, which probably reflects high carriage of ESBL-E. coli in the community. As the warmer and rainy season was the main risk factor associated with ESBL-E. coli carriage, it is reasonable to assume that the situation could grow even worse with climate change trends. Continued surveillance efforts should be maintained along with public health interventions to, if not reverse, at least slow down the spread of resistance in Madagascar. Our results will have to be reconsidered in a One Health context by linking them to the results of the ESBL E. coli prevalence in the food chain and the environment which is the next step of the Tricycle protocol to better understand the complex face of the antimicrobial resistance.

The original contributions presented in the study are publicly available. These data can be found here: https://www.ebi.ac.uk/ena/browser/home (accession number PRJEB48998).

The studies involving human participants were reviewed and approved by the Ethics Committee of Ministry of Health, Madagascar. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin.

MM carried out the experiments, analyzed the results, and wrote the first draft of the manuscript. SR, LVR, RR, RC, CR, ER, and ZR collected the samples and data, and performed the experiments in Madagascar. EW and MP performed the bioinformatic analyses. OC participated in the experiments and analysis of the WGS results. JM performed the statistical analyses. LR, LS, and HE were involved in the implementation of the project in Madagascar and supervised its progress. FK-P managed the implementation of the project and its conduct, and participated in the analysis of the results. AA was involved in the original idea of Tricycle and design of the study. ED was involved in the analysis of the WGS data and revision of the manuscript. LA-L was responsible for the overall direction and planning, supervision of all experiments and results, and writing the manuscript. All authors contributed to the revision of the manuscript and read and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.